Generalized inverse matrix-graphic deep learning algorithm for multispectral pyrometer temperature inversion.

Area of Science:

• Radiometry and Temperature Measurement
• Image Recognition and Deep Learning
• Materials Science and Emissivity Analysis

Background:

• Multispectral radiometric temperature measurement is hindered by unknown emissivity, limiting existing inversion algorithms' applicability.
• A universal algorithm for accurate temperature and emissivity inversion across diverse scenes and targets is currently lacking.

Purpose of the Study:

• To develop a graphical multispectral radiometric temperature adaptive inversion algorithm.
• To address the limitations of unknown emissivity in temperature measurement techniques.
• To enable simultaneous inversion of temperature and emissivity for any scene or target.

Main Methods:

• Converted one-dimensional radiation data into a two-dimensional radiation map using spectral channel differences.
• Utilized generalized inverse to extract spectral emissivity distribution features.
• Fused emissivity features with the radiation map and employed an improved deep learning network for adaptive temperature inversion.

Main Results:

• The proposed graphical algorithm successfully converts radiometric temperature inversion into an image recognition problem.
• Experimental verification demonstrates the algorithm's capability for simultaneous temperature and emissivity inversion.
• The method proves effective for any scene or target, provided a sufficient dataset is available.

Conclusions:

• The developed graphical multispectral radiometric temperature adaptive inversion algorithm offers a robust solution to unknown emissivity challenges.
• This image recognition-based approach significantly enhances the adaptability and accuracy of radiometric temperature measurements.
• The algorithm's ability to perform simultaneous temperature and emissivity inversion opens new possibilities in thermal imaging applications.